A reconfigurable computing architecture based on cellular automata

Abstract

This thesis proposes a reconfigurable computer architecture based on cellular autómata capable of improving computing performance by exploiting the massive parallelism from its individual's interaction. Studies, in the last 15 years, proved that the complexity of the synthesis tools needed to exploit this parallelism increased as the architecture granularity was finer. An architecture based on cellular autómata represents the finest granularity. Studies have proved that cellular autómata granularity can be handle or programmed using genetic algorithms. It is time for a viable reconfigurable computing architecture based on cellular computing to be proposed. The architecture is oriented to exploit nature's parallelism while using the semiconductor technology available nowadays. The philosophy behind is to make the hardware as simple as it may be, and make the software as complex as it is required to be in order to perform valuable computations. An evolutive approach is used to handle the software complexities A road to make this computer architecture feasible is suggested. The first steps towards the implementation of useful cellular automaton computer architecture were explored, including the physical media selection, topology defmition, basic programming tools development, and search for a cell's rule computationally efficient and universal. Examples on how this architecture can compute simple Boolcan functions are presented.